Abstract: With the development of agricultural digitalization and informatization, unmanned equipment-based plant protection is increasingly promoted due to its high efficiency, safety and so on, of which the more widely used are UAV and UGV. To provide more accurate plant protection services, the plant protection trusteeship services providers need to allocate different types of equipment for plant protection. Consequently, taking the perspective of plant protection trusteeship services providers, this paper divides farmland into three types: farmland only operated by UAV, farmland only operated by UGV and farmland where UAV and UGV need to operate synchronously, and studies the path optimization problem of UAV and UGV cooperative pesticide operation in multiple farmlands. The research in this paper has important theoretical significance and practical value. On the one hand, it enriches the theory of unmanned equipment cooperative operation in the context of digital agriculture, and on the other hand, it can provide theoretical support for plant protection custodian service providers in their daily decision-making process.
It is essential for plant protection trusteeship services providers to dispatch several plant protection UAVs as well as plant protection UGVs in the decision-making stage due to the large and multi-type plant protection orders. For the sake of improving the operational efficiency and reducing the cost, it becomes very important to establish an effective optimization method for plant protection operations that fulfill all the orders with the minimum time cost. In this paper, a mixed integer programming model with the optimization objective of minimizing the total operation time is built. Considering the difficulties such as model nonlinearity and multi-stage hybrid decision-making, this paper proposes an iterative multi-stage solving framework that integrates multiple algorithms as follows: firstly, the order of farmland operated by UAV/UGV is derived by ant colony algorithm; based on this order, the simulated annealing algorithm is used to determine the plant protection routes of UAV and UGV operating in each farmland; and then the return point of UAV and UGV is determined; finally, the objective function is solved and the next iteration is performed until a satisfactory solution is obtained. Five groups of farmland data in Heyang County, Weinan City, Shaanxi Province are extracted for simulation calculation. First, we compare the proposed method with existing rules, including the Johnson rule, traditional operation rule, and fixed replenishment rule. The results demonstrate the significant effectiveness of the proposed method in solving the problem. In addition, considering the characteristics of agricultural production and the heterogeneity of plant protection service orders, a sensitivity analysis of the sparsity of farmland distribution and changes in agricultural production scenarios is performed. The conclusions obtained in this paper are as follows: (1)through a comparative analysis with the Johnson rule and traditional operation rule, regardless of the scale of the instances, the method proposed in this paper exhibits significant advantages in both plant protection time costs and non-plant protection time costs; (2)the mobile replenishment mode of trucks considered in this paper saves the number of return times and the distance of return journey of both UAV and UGV, even though this mode increases the energy consumption of the trucks; (3)the distribution of farmland should not be too scattered although UAV and UGV cooperative operation are applicable to plant protection orders with different sparsity, otherwise the efficiency of cooperative operation will be compromised; (4)the waiting time for common service farmland is usually too long because of the need for simultaneous operation of the UAV and UGV, so the number of common service farmland needs to be arranged reasonably.
There are still the following deficiencies in this paper. First, only a single UAV or a single UGV are researched, however, combining multiple UAVs or UGVs into a larger spraying system in reality is a more effective way to improve the operating capability of a single unit of unmanned equipment. Second, the shape of the farmland is a regular rectangle and the obstacles existing in the farmland are ignored, which is a lack of generalizability. All of the above need to be refined in the future research.

Key words: UAV, UGV, collaborative operation optimization, ant colony algorithm, simulated annealing algorithm

摘要： 本文从植保托管服务商角度出发,研究在多作业区域下单无人机与单无人车植保协同作业优化问题。首先构建了以最小化总的作业时间成本为优化目标的混合整数规划模型,考虑到模型非线性、多阶段混合决策等难点,本文提出了混合多个算法的迭代式多阶段求解框架:利用蚁群算法得出无人机与无人车在多个作业区域之间的作业顺序,在此基础上,利用模拟退火算法确定无人机无人车在各农田作业的植保路线,接着,计算无人机与无人车因补药或充电的返航点,最后求解目标函数,进行下一次迭代求解直到获得满意解。实验提取了陕西省渭南市合阳县5组农田数据进行仿真计算,在仿真结果基础上进行了策略对比实验,结果表明,本文所提出的方法具有显著的有效性。此外,考虑到农业生产的特性和植保服务订单的异质性,还进行了农田分布稀疏程度和农业生产场景变化的灵敏度分析。

关键词: 无人机, 无人车, 协同作业优化, 蚁群算法, 模拟退火算法